2ABT Behavior Models

This project contains analysis tools and models of mouse behavior in a two-armed bandit task. The contents of this repo can be used to:

1. Characterize choice and trial-to-trial switching behavior in a 2ABT
2. Model behavior and predict from experimental data
3. Use models to simulate behavior

How to use

Characterize behavior

Code for visualizing choice and switching behavior of animals around block transitions in a dynamic two-armed bandit task as well as for computing and plotting conditional probabilities of behavior given action and outcome history.

Model experimental data

Included models:

1. Hidden Markov model (HMM)
2. Logistic regression
3. Recursively formulated logistic regression (RFLR)
4. forgetting Q-learning model (FQ model)
5. sticky implementation of HMM

Supported action policies:

• Greedy
• "Stochastic"
• Softmax

The notebook demo_models.ipynb demonstrates how to fit and compute choice probabilities using the various models. Mouse data analyzed in Beron et al., 2022 can be found at https://doi.org/10.7910/DVN/7E0NM5 (note, this has changed from previous location).

Simulate behavior from models

This repo currently includes generative simulations for the HMM and RFLR.

Installation

git clone https://github.com/celiaberon/2ABT_models
cd 2ABT_models
conda create -n 2abt-models python=3.8
conda activate 2abt-models
pip install -r requirements.txt

After setting up the virtual environment, install the SSM package for building and using the HMM following the instructions at https://github.com/lindermanlab/ssm.